def test_linearsvc_parameters():
# Test possible parameter combinations in LinearSVC
# Generate list of possible parameter combinations
losses = ['hinge', 'squared_hinge', 'logistic_regression', 'foo']
penalties, duals = ['l1', 'l2', 'bar'], [True, False]
X, y = make_classification(n_samples=5, n_features=5)
for loss, penalty, dual in itertools.product(losses, penalties, duals):
clf = svm.LinearSVC(penalty=penalty, loss=loss, dual=dual)
if ((loss, penalty) == ('hinge', 'l1') or
(loss, penalty, dual) == ('hinge', 'l2', False) or
(penalty, dual) == ('l1', True) or
loss == 'foo' or penalty == 'bar'):
assert_raises_regexp(ValueError,
"Unsupported set of arguments.*penalty='%s.*"
"loss='%s.*dual=%s"
% (penalty, loss, dual),
clf.fit, X, y)
else:
clf.fit(X, y)
# Incorrect loss value - test if explicit error message is raised
assert_raises_regexp(ValueError, ".*loss='l3' is not supported.*",
svm.LinearSVC(loss="l3").fit, X, y)
def test_non_positive_precomputed_distances():
# Precomputed distance matrices must be positive.
bad_dist = np.array([[0., -1.], [1., 0.]])
for method in ['barnes_hut', 'exact']:
tsne = TSNE(metric="precomputed", method=method)
assert_raises_regexp(ValueError, "All distances .*precomputed.*",
tsne.fit_transform, bad_dist)
def test_check_scoring():
# Test all branches of check_scoring
estimator = EstimatorWithoutFit()
pattern = (r"estimator should a be an estimator implementing 'fit' method,"
r" .* was passed")
assert_raises_regexp(TypeError, pattern, check_scoring, estimator)
estimator = EstimatorWithFitAndScore()
estimator.fit([[1]], [1])
scorer = check_scoring(estimator)
assert_true(scorer is _passthrough_scorer)
assert_almost_equal(scorer(estimator, [[1]], [1]), 1.0)
estimator = EstimatorWithFitAndPredict()
estimator.fit([[1]], [1])
pattern = (r"If no scoring is specified, the estimator passed should have"
r" a 'score' method\. The estimator .* does not\.")
assert_raises_regexp(TypeError, pattern, check_scoring, estimator)
scorer = check_scoring(estimator, "accuracy")
assert_almost_equal(scorer(estimator, [[1]], [1]), 1.0)
estimator = EstimatorWithFit()
scorer = check_scoring(estimator, "accuracy")
assert_true(isinstance(scorer, _PredictScorer))
estimator = EstimatorWithFit()
scorer = check_scoring(estimator, allow_none=True)
assert_true(scorer is None)
def check_scoring_validator_for_single_metric_usecases(scoring_validator):
# Test all branches of single metric usecases
estimator = EstimatorWithoutFit()
pattern = (r"estimator should be an estimator implementing 'fit' method,"
r" .* was passed")
assert_raises_regexp(TypeError, pattern, scoring_validator, estimator)
estimator = EstimatorWithFitAndScore()
estimator.fit([[1]], [1])
scorer = scoring_validator(estimator)
assert_true(scorer is _passthrough_scorer)
assert_almost_equal(scorer(estimator, [[1]], [1]), 1.0)
estimator = EstimatorWithFitAndPredict()
estimator.fit([[1]], [1])
pattern = (r"If no scoring is specified, the estimator passed should have"
r" a 'score' method\. The estimator .* does not\.")
assert_raises_regexp(TypeError, pattern, scoring_validator, estimator)
scorer = scoring_validator(estimator, "accuracy")
assert_almost_equal(scorer(estimator, [[1]], [1]), 1.0)
estimator = EstimatorWithFit()
scorer = scoring_validator(estimator, "accuracy")
assert_true(isinstance(scorer, _PredictScorer))
# Test the allow_none parameter for check_scoring alone
if scoring_validator is check_scoring:
estimator = EstimatorWithFit()
scorer = scoring_validator(estimator, allow_none=True)
assert_true(scorer is None)
def test_ransac_resid_thresh_no_inliers():
# When residual_threshold=0.0 there are no inliers and a
# ValueError with a message should be raised
base_estimator = LinearRegression()
ransac_estimator = RANSACRegressor(base_estimator, min_samples=2, residual_threshold=0.0, random_state=0)
assert_raises_regexp(ValueError, "No inliers.*residual_threshold.*0\.0", ransac_estimator.fit, X, y)
def test_base_estimator():
# Test different base estimators.
from sklearn.ensemble import RandomForestClassifier
from sklearn.svm import SVC
# XXX doesn't work with y_class because RF doesn't support classes_
# Shouldn't AdaBoost run a LabelBinarizer?
clf = AdaBoostClassifier(RandomForestClassifier())
clf.fit(X, y_regr)
clf = AdaBoostClassifier(SVC(), algorithm="SAMME")
clf.fit(X, y_class)
from sklearn.ensemble import RandomForestRegressor
from sklearn.svm import SVR
clf = AdaBoostRegressor(RandomForestRegressor(), random_state=0)
clf.fit(X, y_regr)
clf = AdaBoostRegressor(SVR(), random_state=0)
clf.fit(X, y_regr)
# Check that an empty discrete ensemble fails in fit, not predict.
X_fail = [[1, 1], [1, 1], [1, 1], [1, 1]]
y_fail = ["foo", "bar", 1, 2]
clf = AdaBoostClassifier(SVC(), algorithm="SAMME")
assert_raises_regexp(ValueError, "worse than random",
clf.fit, X_fail, y_fail)
def test_underflow_or_overlow():
with np.errstate(all="raise"):
# Generate some weird data with hugely unscaled features
rng = np.random.RandomState(0)
n_samples = 100
n_features = 10
X = rng.normal(size=(n_samples, n_features))
X[:, :2] *= 1e300
assert_true(np.isfinite(X).all())
# Use MinMaxScaler to scale the data without introducing a numerical
# instability (computing the standard deviation naively is not possible
# on this data)
X_scaled = MinMaxScaler().fit_transform(X)
assert_true(np.isfinite(X_scaled).all())
# Define a ground truth on the scaled data
ground_truth = rng.normal(size=n_features)
y = (np.dot(X_scaled, ground_truth) > 0.0).astype(np.int32)
assert_array_equal(np.unique(y), [0, 1])
model = SGDClassifier(alpha=0.1, loss="squared_hinge", n_iter=500)
# smoke test: model is stable on scaled data
model.fit(X_scaled, y)
assert_true(np.isfinite(model.coef_).all())
# model is numerically unstable on unscaled data
msg_regxp = (
r"Floating-point under-/overflow occurred at epoch #.*"
" Scaling input data with StandardScaler or MinMaxScaler"
" might help."
)
assert_raises_regexp(ValueError, msg_regxp, model.fit, X, y)
def test_lda_no_component_error():
# test `perplexity` before `fit`
rng = np.random.RandomState(0)
X = rng.randint(4, size=(20, 10))
lda = LatentDirichletAllocation()
regex = ("This LatentDirichletAllocation instance is not fitted yet. "
"Call 'fit' with appropriate arguments before using this method.")
assert_raises_regexp(NotFittedError, regex, lda.perplexity, X)
def test_no_sparse_on_barnes_hut():
# No sparse matrices allowed on Barnes-Hut.
random_state = check_random_state(0)
X = random_state.randn(100, 2)
X[(np.random.randint(0, 100, 50), np.random.randint(0, 2, 50))] = 0.0
X_csr = sp.csr_matrix(X)
tsne = TSNE(n_iter=199, method="barnes_hut")
assert_raises_regexp(TypeError, "A sparse matrix was.*", tsne.fit_transform, X_csr)
def test_unfitted():
X = "foo!" # input validation not required when SVM not fitted
clf = svm.SVC()
assert_raises_regexp(Exception, r".*\bSVC\b.*\bnot\b.*\bfitted\b", clf.predict, X)
clf = svm.NuSVR()
assert_raises_regexp(Exception, r".*\bNuSVR\b.*\bnot\b.*\bfitted\b", clf.predict, X)
def test_k_means_n_init():
rnd = np.random.RandomState(0)
X = rnd.normal(size=(40, 2))
# two regression tests on bad n_init argument
# previous bug: n_init <= 0 threw non-informative TypeError (#3858)
assert_raises_regexp(ValueError, "n_init", KMeans(n_init=0).fit, X)
assert_raises_regexp(ValueError, "n_init", KMeans(n_init=-1).fit, X)
def test_lda_no_component_error():
# test `transform` and `perplexity` before `fit`
rng = np.random.RandomState(0)
X = rng.randint(4, size=(20, 10))
lda = LatentDirichletAllocation()
regex = r"^no 'components_' attribute"
assert_raises_regexp(NotFittedError, regex, lda.transform, X)
assert_raises_regexp(NotFittedError, regex, lda.perplexity, X)
def test_distance_not_available():
# 'metric' must be valid.
tsne = TSNE(metric="not available", method='exact')
assert_raises_regexp(ValueError, "Unknown metric not available.*",
tsne.fit_transform, np.array([[0.0], [1.0]]))
tsne = TSNE(metric="not available", method='barnes_hut')
assert_raises_regexp(ValueError, "Metric 'not available' not valid.*",
tsne.fit_transform, np.array([[0.0], [1.0]]))
def test_non_positive_computed_distances():
# Computed distance matrices must be positive.
def metric(x, y):
return -1
tsne = TSNE(metric=metric, method='exact')
X = np.array([[0.0, 0.0], [1.0, 1.0]])
assert_raises_regexp(ValueError, "All distances .*metric given.*",
tsne.fit_transform, X)
def test_pca_initialization_not_compatible_with_precomputed_kernel():
# Precomputed distance matrices must be square matrices.
tsne = TSNE(metric="precomputed", init="pca")
assert_raises_regexp(
ValueError,
'The parameter init="pca" cannot be ' 'used with metric="precomputed".',
tsne.fit_transform,
np.array([[0.0], [1.0]]),
)
def test_lda_transform_mismatch():
# test `n_features` mismatch in partial_fit and transform
rng = np.random.RandomState(0)
X = rng.randint(4, size=(20, 10))
X_2 = rng.randint(4, size=(10, 8))
n_topics = rng.randint(3, 6)
lda = LatentDirichletAllocation(n_topics=n_topics, random_state=rng)
lda.partial_fit(X)
assert_raises_regexp(ValueError, r"^The provided data has", lda.partial_fit, X_2)
def test_ovr_partial_fit_exceptions():
ovr = OneVsRestClassifier(MultinomialNB())
X = np.abs(np.random.randn(14, 2))
y = [1, 1, 1, 1, 2, 3, 3, 0, 0, 2, 3, 1, 2, 3]
ovr.partial_fit(X[:7], y[:7], np.unique(y))
# A new class value which was not in the first call of partial_fit
# It should raise ValueError
y1 = [5] + y[7:-1]
assert_raises_regexp(ValueError, r"Mini-batch contains \[.+\] while "
r"classes must be subset of \[.+\]",
ovr.partial_fit, X=X[7:], y=y1)
def test_lda_partial_fit_dim_mismatch():
# test `n_features` mismatch in `partial_fit`
rng = np.random.RandomState(0)
n_topics = rng.randint(3, 6)
n_col = rng.randint(6, 10)
X_1 = np.random.randint(4, size=(10, n_col))
X_2 = np.random.randint(4, size=(10, n_col + 1))
lda = LatentDirichletAllocation(n_topics=n_topics, learning_offset=5.,
total_samples=20, random_state=rng)
lda.partial_fit(X_1)
assert_raises_regexp(ValueError, r"^The provided data has", lda.partial_fit, X_2)
def test_invalid_params():
# test `_check_params` method
X = np.ones((5, 10))
invalid_models = (
('n_topics', LatentDirichletAllocation(n_topics=0)),
('learning_method', LatentDirichletAllocation(learning_method='unknown')),
('total_samples', LatentDirichletAllocation(total_samples=0)),
('learning_offset', LatentDirichletAllocation(learning_offset=-1)),
)
for param, model in invalid_models:
regex = r"^Invalid %r parameter" % param
assert_raises_regexp(ValueError, regex, model.fit, X)
def test_ransac_resid_thresh_no_inliers():
# When residual_threshold=0.0 there are no inliers and a
# ValueError with a message should be raised
base_estimator = LinearRegression()
ransac_estimator = RANSACRegressor(base_estimator, min_samples=2,
residual_threshold=0.0, random_state=0,
max_trials=5)
msg = ("RANSAC could not find a valid consensus set")
assert_raises_regexp(ValueError, msg, ransac_estimator.fit, X, y)
assert_equal(ransac_estimator.n_skips_no_inliers_, 5)
assert_equal(ransac_estimator.n_skips_invalid_data_, 0)
assert_equal(ransac_estimator.n_skips_invalid_model_, 0)
def test_partial_fit_weight_class_auto(self):
# partial_fit with class_weight='auto' not supported
assert_raises_regexp(ValueError,
"class_weight 'auto' is not supported for "
"partial_fit. In order to use 'auto' weights, "
"use compute_class_weight\('auto', classes, y\). "
"In place of y you can us a large enough sample "
"of the full training set target to properly "
"estimate the class frequency distributions. "
"Pass the resulting weights as the class_weight "
"parameter.",
self.factory(class_weight='auto').partial_fit,
X, Y, classes=np.unique(Y))
def test_ransac_no_valid_model():
def is_model_valid(estimator, X, y):
return False
base_estimator = LinearRegression()
ransac_estimator = RANSACRegressor(base_estimator,
is_model_valid=is_model_valid,
max_trials=5)
msg = ("RANSAC could not find a valid consensus set")
assert_raises_regexp(ValueError, msg, ransac_estimator.fit, X, y)
assert_equal(ransac_estimator.n_skips_no_inliers_, 0)
assert_equal(ransac_estimator.n_skips_invalid_data_, 0)
assert_equal(ransac_estimator.n_skips_invalid_model_, 5)
def test_ovo_partial_fit_predict():
temp = datasets.load_iris()
X, y = temp.data, temp.target
ovo1 = OneVsOneClassifier(MultinomialNB())
ovo1.partial_fit(X[:100], y[:100], np.unique(y))
ovo1.partial_fit(X[100:], y[100:])
pred1 = ovo1.predict(X)
ovo2 = OneVsOneClassifier(MultinomialNB())
ovo2.fit(X, y)
pred2 = ovo2.predict(X)
assert_equal(len(ovo1.estimators_), n_classes * (n_classes - 1) / 2)
assert_greater(np.mean(y == pred1), 0.65)
assert_almost_equal(pred1, pred2)
# Test when mini-batches have binary target classes
ovo1 = OneVsOneClassifier(MultinomialNB())
ovo1.partial_fit(X[:60], y[:60], np.unique(y))
ovo1.partial_fit(X[60:], y[60:])
pred1 = ovo1.predict(X)
ovo2 = OneVsOneClassifier(MultinomialNB())
pred2 = ovo2.fit(X, y).predict(X)
assert_almost_equal(pred1, pred2)
assert_equal(len(ovo1.estimators_), len(np.unique(y)))
assert_greater(np.mean(y == pred1), 0.65)
ovo = OneVsOneClassifier(MultinomialNB())
X = np.random.rand(14, 2)
y = [1, 1, 2, 3, 3, 0, 0, 4, 4, 4, 4, 4, 2, 2]
ovo.partial_fit(X[:7], y[:7], [0, 1, 2, 3, 4])
ovo.partial_fit(X[7:], y[7:])
pred = ovo.predict(X)
ovo2 = OneVsOneClassifier(MultinomialNB())
pred2 = ovo2.fit(X, y).predict(X)
assert_almost_equal(pred, pred2)
# raises error when mini-batch does not have classes from all_classes
ovo = OneVsOneClassifier(MultinomialNB())
error_y = [0, 1, 2, 3, 4, 5, 2]
message_re = escape("Mini-batch contains {0} while "
"it must be subset of {1}".format(np.unique(error_y),
np.unique(y)))
assert_raises_regexp(ValueError, message_re, ovo.partial_fit, X[:7],
error_y, np.unique(y))
# test partial_fit only exists if estimator has it:
ovr = OneVsOneClassifier(SVC())
assert_false(hasattr(ovr, "partial_fit"))
def test_ransac_exceed_max_skips():
def is_data_valid(X, y):
return False
base_estimator = LinearRegression()
ransac_estimator = RANSACRegressor(base_estimator,
is_data_valid=is_data_valid,
max_trials=5,
max_skips=3)
msg = ("RANSAC skipped more iterations than `max_skips`")
assert_raises_regexp(ValueError, msg, ransac_estimator.fit, X, y)
assert_equal(ransac_estimator.n_skips_no_inliers_, 0)
assert_equal(ransac_estimator.n_skips_invalid_data_, 4)
assert_equal(ransac_estimator.n_skips_invalid_model_, 0)
def test_lda_preplexity_mismatch():
# test dimension mismatch in `perplexity` method
rng = np.random.RandomState(0)
n_topics = rng.randint(3, 6)
n_samples = rng.randint(6, 10)
X = np.random.randint(4, size=(n_samples, 10))
lda = LatentDirichletAllocation(n_topics=n_topics, learning_offset=5.,
total_samples=20, random_state=rng)
lda.fit(X)
# invalid samples
invalid_n_samples = rng.randint(4, size=(n_samples + 1, n_topics))
assert_raises_regexp(ValueError, r'Number of samples', lda.perplexity, X, invalid_n_samples)
# invalid topic number
invalid_n_topics = rng.randint(4, size=(n_samples, n_topics + 1))
assert_raises_regexp(ValueError, r'Number of topics', lda.perplexity, X, invalid_n_topics)
def test_scoring_is_not_metric():
assert_raises_regexp(ValueError, 'make_scorer', check_scoring,
LogisticRegression(), f1_score)
assert_raises_regexp(ValueError, 'make_scorer', check_scoring,
LogisticRegression(), roc_auc_score)
assert_raises_regexp(ValueError, 'make_scorer', check_scoring,
Ridge(), r2_score)
assert_raises_regexp(ValueError, 'make_scorer', check_scoring,
KMeans(), cluster_module.adjusted_rand_score)
def test_check_scoring_and_check_multimetric_scoring():
check_scoring_validator_for_single_metric_usecases(check_scoring)
# To make sure the check_scoring is correctly applied to the constituent
# scorers
check_scoring_validator_for_single_metric_usecases(
check_multimetric_scoring_single_metric_wrapper)
# For multiple metric use cases
# Make sure it works for the valid cases
for scoring in (('accuracy',), ['precision'],
{'acc': 'accuracy', 'precision': 'precision'},
('accuracy', 'precision'), ['precision', 'accuracy'],
{'accuracy': make_scorer(accuracy_score),
'precision': make_scorer(precision_score)}):
estimator = LinearSVC(random_state=0)
estimator.fit([[1], [2], [3]], [1, 1, 0])
scorers, is_multi = _check_multimetric_scoring(estimator, scoring)
assert_true(is_multi)
assert_true(isinstance(scorers, dict))
assert_equal(sorted(scorers.keys()), sorted(list(scoring)))
assert_true(all([isinstance(scorer, _PredictScorer)
for scorer in list(scorers.values())]))
if 'acc' in scoring:
assert_almost_equal(scorers['acc'](
estimator, [[1], [2], [3]], [1, 0, 0]), 2. / 3.)
if 'accuracy' in scoring:
assert_almost_equal(scorers['accuracy'](
estimator, [[1], [2], [3]], [1, 0, 0]), 2. / 3.)
if 'precision' in scoring:
assert_almost_equal(scorers['precision'](
estimator, [[1], [2], [3]], [1, 0, 0]), 0.5)
estimator = EstimatorWithFitAndPredict()
estimator.fit([[1]], [1])
# Make sure it raises errors when scoring parameter is not valid.
# More weird corner cases are tested at test_validation.py
error_message_regexp = ".*must be unique strings.*"
for scoring in ((make_scorer(precision_score), # Tuple of callables
make_scorer(accuracy_score)), [5],
(make_scorer(precision_score),), (), ('f1', 'f1')):
assert_raises_regexp(ValueError, error_message_regexp,
_check_multimetric_scoring, estimator,
scoring=scoring)
def test_correct_labelsize():
# Assert 1 < n_labels < n_samples
dataset = datasets.load_iris()
X = dataset.data
# n_labels = n_samples
y = np.arange(X.shape[0])
assert_raises_regexp(ValueError,
'Number of labels is %d\. Valid values are 2 '
'to n_samples - 1 \(inclusive\)' % len(np.unique(y)),
silhouette_score, X, y)
# n_labels = 1
y = np.zeros(X.shape[0])
assert_raises_regexp(ValueError,
'Number of labels is %d\. Valid values are 2 '
'to n_samples - 1 \(inclusive\)' % len(np.unique(y)),
silhouette_score, X, y)
def test_correct_labelsize():
""" Assert 2 <= n_labels <= nsample -1 """
dataset = datasets.load_iris()
X = dataset.data
# n_labels = n_samples
y = np.arange(X.shape[0])
assert_raises_regexp(ValueError,
"Number of labels is %d "
"but should be more than 2"
"and less than n_samples - 1" % len(np.unique(y)),
silhouette_score, X, y)
# n_labels = 1
y = np.zeros(X.shape[0])
assert_raises_regexp(ValueError,
"Number of labels is %d "
"but should be more than 2"
"and less than n_samples - 1" % len(np.unique(y)),
silhouette_score, X, y)
def test_pairwise_precomputed(func):
# Test correct shape
assert_raises_regexp(ValueError, '.* shape .*',
func, np.zeros((5, 3)), metric='precomputed')
# with two args
assert_raises_regexp(ValueError, '.* shape .*',
func, np.zeros((5, 3)), np.zeros((4, 4)),
metric='precomputed')
# even if shape[1] agrees (although thus second arg is spurious)
assert_raises_regexp(ValueError, '.* shape .*',
func, np.zeros((5, 3)), np.zeros((4, 3)),
metric='precomputed')
# Test not copied (if appropriate dtype)
S = np.zeros((5, 5))
S2 = func(S, metric="precomputed")
assert S is S2
# with two args
S = np.zeros((5, 3))
S2 = func(S, np.zeros((3, 3)), metric="precomputed")
assert S is S2
# Test always returns float dtype
S = func(np.array([[1]], dtype='int'), metric='precomputed')
assert_equal('f', S.dtype.kind)
# Test converts list to array-like
S = func([[1.]], metric='precomputed')
assert isinstance(S, np.ndarray)
def test_pairwise_precomputed():
for func in [pairwise_distances, pairwise_kernels]:
# Test correct shape
assert_raises_regexp(ValueError,
'.* shape .*',
func,
np.zeros((5, 3)),
metric='precomputed')
# with two args
assert_raises_regexp(ValueError,
'.* shape .*',
func,
np.zeros((5, 3)),
np.zeros((4, 4)),
metric='precomputed')
# even if shape[1] agrees (although thus second arg is spurious)
assert_raises_regexp(ValueError,
'.* shape .*',
func,
np.zeros((5, 3)),
np.zeros((4, 3)),
metric='precomputed')
# Test not copied (if appropriate dtype)
S = np.zeros((5, 5))
S2 = func(S, metric="precomputed")
assert_true(S is S2)
# with two args
S = np.zeros((5, 3))
S2 = func(S, np.zeros((3, 3)), metric="precomputed")
assert_true(S is S2)
# Test always returns float dtype
S = func(np.array([[1]], dtype='int'), metric='precomputed')
assert_equal('f', S.dtype.kind)
# Test converts list to array-like
S = func([[1]], metric='precomputed')
assert_true(isinstance(S, np.ndarray))
def test_init_not_available():
# 'init' must be 'pca', 'random', or numpy array.
tsne = TSNE(init="not available")
m = "'init' must be 'pca', 'random', or a numpy array"
assert_raises_regexp(ValueError, m, tsne.fit_transform,
np.array([[0.0], [1.0]]))
def test_group_kfold():
rng = np.random.RandomState(0)
# Parameters of the test
n_groups = 15
n_samples = 1000
n_splits = 5
X = y = np.ones(n_samples)
# Construct the test data
tolerance = 0.05 * n_samples # 5 percent error allowed
groups = rng.randint(0, n_groups, n_samples)
ideal_n_groups_per_fold = n_samples // n_splits
len(np.unique(groups))
# Get the test fold indices from the test set indices of each fold
folds = np.zeros(n_samples)
lkf = GroupKFold(n_splits=n_splits)
for i, (_, test) in enumerate(lkf.split(X, y, groups)):
folds[test] = i
# Check that folds have approximately the same size
assert_equal(len(folds), len(groups))
for i in np.unique(folds):
assert_greater_equal(tolerance,
abs(sum(folds == i) - ideal_n_groups_per_fold))
# Check that each group appears only in 1 fold
for group in np.unique(groups):
assert_equal(len(np.unique(folds[groups == group])), 1)
# Check that no group is on both sides of the split
groups = np.asarray(groups, dtype=object)
for train, test in lkf.split(X, y, groups):
assert_equal(len(np.intersect1d(groups[train], groups[test])), 0)
# Construct the test data
groups = np.array([
'Albert', 'Jean', 'Bertrand', 'Michel', 'Jean', 'Francis', 'Robert',
'Michel', 'Rachel', 'Lois', 'Michelle', 'Bernard', 'Marion', 'Laura',
'Jean', 'Rachel', 'Franck', 'John', 'Gael', 'Anna', 'Alix', 'Robert',
'Marion', 'David', 'Tony', 'Abel', 'Becky', 'Madmood', 'Cary', 'Mary',
'Alexandre', 'David', 'Francis', 'Barack', 'Abdoul', 'Rasha', 'Xi',
'Silvia'
])
n_groups = len(np.unique(groups))
n_samples = len(groups)
n_splits = 5
tolerance = 0.05 * n_samples # 5 percent error allowed
ideal_n_groups_per_fold = n_samples // n_splits
X = y = np.ones(n_samples)
# Get the test fold indices from the test set indices of each fold
folds = np.zeros(n_samples)
for i, (_, test) in enumerate(lkf.split(X, y, groups)):
folds[test] = i
# Check that folds have approximately the same size
assert_equal(len(folds), len(groups))
for i in np.unique(folds):
assert_greater_equal(tolerance,
abs(sum(folds == i) - ideal_n_groups_per_fold))
# Check that each group appears only in 1 fold
with warnings.catch_warnings():
warnings.simplefilter("ignore", DeprecationWarning)
for group in np.unique(groups):
assert_equal(len(np.unique(folds[groups == group])), 1)
# Check that no group is on both sides of the split
groups = np.asarray(groups, dtype=object)
for train, test in lkf.split(X, y, groups):
assert_equal(len(np.intersect1d(groups[train], groups[test])), 0)
# groups can also be a list
cv_iter = list(lkf.split(X, y, groups.tolist()))
for (train1, test1), (train2, test2) in zip(lkf.split(X, y, groups),
cv_iter):
assert_array_equal(train1, train2)
assert_array_equal(test1, test2)
# Should fail if there are more folds than groups
groups = np.array([1, 1, 1, 2, 2])
X = y = np.ones(len(groups))
assert_raises_regexp(ValueError, "Cannot have number of splits.*greater",
next,
GroupKFold(n_splits=3).split(X, y, groups))
def test_init_not_available():
# 'init' must be 'pca' or 'random'.
assert_raises_regexp(ValueError,
"'init' must be either 'pca' or 'random'",
TSNE,
init="not available")
def test_non_square_precomputed_distances():
# Precomputed distance matrices must be square matrices.
tsne = TSNE(metric="precomputed")
assert_raises_regexp(ValueError, ".* square distance matrix",
tsne.fit_transform, np.array([[0.0], [1.0]]))
def test_too_few_iterations():
# Number of gradient descent iterations must be at least 200.
tsne = TSNE(n_iter=199)
assert_raises_regexp(ValueError, "n_iter .*", tsne.fit_transform,
np.array([[0.0], [0.0]]))
def test_pairwise_precomputed_non_negative():
# Test non-negative values
assert_raises_regexp(ValueError, '.* non-negative values.*',
pairwise_distances, np.full((5, 5), -1),
metric='precomputed')
def test_distance_not_available():
# 'metric' must be valid.
tsne = TSNE(metric="not available")
assert_raises_regexp(ValueError, "Unknown metric not available.*",
tsne.fit_transform, np.array([[0.0], [1.0]]))
def test_init_not_available():
# 'init' must be 'pca' or 'random'.
m = "'init' must be 'pca', 'random' or a NumPy array"
assert_raises_regexp(ValueError, m, TSNE, init="not available")
def test_pairwise_distances_chunked_reduce_invalid(bad_reduce, err_type,
message):
X = np.arange(10).reshape(-1, 1)
S_chunks = pairwise_distances_chunked(X, None, reduce_func=bad_reduce,
working_memory=64)
assert_raises_regexp(err_type, message, next, S_chunks)
def test_label_kfold():
rng = np.random.RandomState(0)
# Parameters of the test
n_labels = 15
n_samples = 1000
n_folds = 5
X = y = np.ones(n_samples)
# Construct the test data
tolerance = 0.05 * n_samples # 5 percent error allowed
labels = rng.randint(0, n_labels, n_samples)
ideal_n_labels_per_fold = n_samples // n_folds
len(np.unique(labels))
# Get the test fold indices from the test set indices of each fold
folds = np.zeros(n_samples)
lkf = LabelKFold(n_folds=n_folds)
for i, (_, test) in enumerate(lkf.split(X, y, labels)):
folds[test] = i
# Check that folds have approximately the same size
assert_equal(len(folds), len(labels))
for i in np.unique(folds):
assert_greater_equal(tolerance,
abs(sum(folds == i) - ideal_n_labels_per_fold))
# Check that each label appears only in 1 fold
for label in np.unique(labels):
assert_equal(len(np.unique(folds[labels == label])), 1)
# Check that no label is on both sides of the split
labels = np.asarray(labels, dtype=object)
for train, test in lkf.split(X, y, labels):
assert_equal(len(np.intersect1d(labels[train], labels[test])), 0)
# Construct the test data
labels = [
'Albert', 'Jean', 'Bertrand', 'Michel', 'Jean', 'Francis', 'Robert',
'Michel', 'Rachel', 'Lois', 'Michelle', 'Bernard', 'Marion', 'Laura',
'Jean', 'Rachel', 'Franck', 'John', 'Gael', 'Anna', 'Alix', 'Robert',
'Marion', 'David', 'Tony', 'Abel', 'Becky', 'Madmood', 'Cary', 'Mary',
'Alexandre', 'David', 'Francis', 'Barack', 'Abdoul', 'Rasha', 'Xi',
'Silvia'
]
n_labels = len(np.unique(labels))
n_samples = len(labels)
n_folds = 5
tolerance = 0.05 * n_samples # 5 percent error allowed
ideal_n_labels_per_fold = n_samples // n_folds
X = y = np.ones(n_samples)
# Get the test fold indices from the test set indices of each fold
folds = np.zeros(n_samples)
for i, (_, test) in enumerate(lkf.split(X, y, labels)):
folds[test] = i
# Check that folds have approximately the same size
assert_equal(len(folds), len(labels))
for i in np.unique(folds):
assert_greater_equal(tolerance,
abs(sum(folds == i) - ideal_n_labels_per_fold))
# Check that each label appears only in 1 fold
with warnings.catch_warnings():
warnings.simplefilter("ignore", DeprecationWarning)
for label in np.unique(labels):
assert_equal(len(np.unique(folds[labels == label])), 1)
# Check that no label is on both sides of the split
labels = np.asarray(labels, dtype=object)
for train, test in lkf.split(X, y, labels):
assert_equal(len(np.intersect1d(labels[train], labels[test])), 0)
# Should fail if there are more folds than labels
labels = np.array([1, 1, 1, 2, 2])
X = y = np.ones(len(labels))
assert_raises_regexp(ValueError, "Cannot have number of folds.*greater",
next,
LabelKFold(n_folds=3).split(X, y, labels))
def test_pca_initialization_not_compatible_with_precomputed_kernel():
# Precomputed distance matrices must be square matrices.
tsne = TSNE(metric="precomputed", init="pca")
assert_raises_regexp(ValueError, "The parameter init=\"pca\" cannot be "
"used with metric=\"precomputed\".",
tsne.fit_transform, np.array([[0.0], [1.0]]))
def test_method_not_available():
# 'nethod' must be 'barnes_hut' or 'exact'
tsne = TSNE(method='not available')
assert_raises_regexp(ValueError, "'method' must be 'barnes_hut' or ",
tsne.fit_transform, np.array([[0.0], [1.0]]))
def test_early_exaggeration_too_small():
# Early exaggeration factor must be >= 1.
tsne = TSNE(early_exaggeration=0.99)
assert_raises_regexp(ValueError, "early_exaggeration .*",
tsne.fit_transform, np.array([[0.0], [0.0]]))
def test_angle_out_of_range_checks():
# check the angle parameter range
for angle in [-1, -1e-6, 1 + 1e-6, 2]:
tsne = TSNE(angle=angle)
assert_raises_regexp(ValueError, "'angle' must be between 0.0 - 1.0",
tsne.fit_transform, np.array([[0.0], [1.0]]))
def test_lda_negative_input():
# test pass dense matrix with sparse negative input.
X = -np.ones((5, 10))
lda = LatentDirichletAllocation()
regex = r"^Negative values in data passed"
assert_raises_regexp(ValueError, regex, lda.fit, X)
def test_n_components_range():
# barnes_hut method should only be used with n_components <= 3
tsne = TSNE(n_components=4, method="barnes_hut")
assert_raises_regexp(ValueError, "'n_components' should be .*",
tsne.fit_transform, np.array([[0.0], [1.0]]))
def test_error():
# Test for appropriate exception on errors
msg = "Penalty term must be positive"
assert_raises_regexp(ValueError, msg, LogisticRegression(C=-1).fit, X, Y1)
assert_raises_regexp(ValueError, msg,
LogisticRegression(C="test").fit, X, Y1)
msg = "Tolerance for stopping criteria must be positive"
assert_raises_regexp(ValueError, msg,
LogisticRegression(tol=-1).fit, X, Y1)
assert_raises_regexp(ValueError, msg,
LogisticRegression(tol="test").fit, X, Y1)
msg = "Maximum number of iteration must be positive"
assert_raises_regexp(ValueError, msg,
LogisticRegression(max_iter=-1).fit, X, Y1)
assert_raises_regexp(ValueError, msg,
LogisticRegression(max_iter="test").fit, X, Y1)